Decomposition of wage losses due to job displacement Pedro Raposo - - PowerPoint PPT Presentation

SLIDE 1

Decomposition of wage losses due to job displacement

Pedro Raposo (Catolica Lisbon SBE, Universidade Catolica Portuguesa) Pedro Portugal (Banco de Portugal, NOVA SBE and IZA Bonn) Anabela Carneiro (Universidade do Porto and CETE)

Paris, OECD, May 2013

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 1 / 21

SLIDE 2

Introduction

Monthly earnings of workers separating in year 2002 and non-displaced workers

0 ¡ 200 ¡ 400 ¡ 600 ¡ 800 ¡ 1000 ¡ 1200 ¡ 1400 ¡ 1600 ¡ 1800 ¡ 2000 ¡ 1997 ¡ 1998 ¡ 1999 ¡ 2000 ¡ 2002 ¡ 2003 ¡ 2004 ¡ 2005 ¡ 2006 ¡ 2007 ¡ 2008 ¡

Non-­‑displaced ¡ Firm ¡closure ¡ Collec=ve ¡dismissals ¡ Individual ¡dismissals ¡

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 2 / 21

SLIDE 3

Introduction

Motivation

Displaced workers experience substantial and persistent reductions in earnings - on the order of 8 to 25 percent for prime-aged workers, in comparison with their non-displaced counterparts (Couch and Placzek, 2010) lasting over 15-20 years (Wachter, 2010). The contribution of human capital to the wage growth has been decomposed in several components

general human capital firm specific human capital job (or task)-specific human capital.

Explain the sources of wage losses

Wage policy of firms Job title assignment Selection into employment Accounting for idiosyncratic trends

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 3 / 21

SLIDE 4

Introduction

Worker (constant) heterogeneity

Worker heterogeneity and the Diamond’s paradox Worker observed permanent characteristics:

Gender and race Formal education (if it does not change over the working life) Birth cohort

Worker unobserved permanent characteristics

Ability education quality Family background Employment cohort (which may be unobserved) Risk aversion Colour of the eyes Beauty (assuming it does not change over the sample period...Heroic!) DNA (which of, course, includes gender, race, eyes’s coulor, freckles, etc.)

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 4 / 21

SLIDE 5

Introduction

Wage policy heterogeneity among firms

Theories that explain why firms find it profitable to pay non-competitive wages

Implicit contracts, principal-agent, efficiency-wages, rent-sharing, and insider-outsider considerations

Firms design incentive schemes to retain their workers, attract better workers, and enhance their productivity (compensation and retention policies) Labor market frictions explanations for the wage differentials: job search and matching literature Permanent characteristics of firms: Location and industry Managerial ability and managerial organization Hiring and firing policies ...as long they do not change over the sample period

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 5 / 21

SLIDE 6

Introduction

Job title heterogeneity

Job title heterogeneity has been neglected...at most some, attention has been placed on the role of occupations There are compensating differentials for certain occupations involving:

Risks of accidents/injuries Stressful working conditions Complexity of tasks (requiring specific training or unusual skills)

Unions may limit or speed up the accession to job-titles

Closed shops Promotion policies Definition of job-titles (and its level of stratification)

Persistent oversupply of labor for some occupations (teachers)

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 6 / 21

SLIDE 7

Introduction

Previous Literature

Addison and Portugal (1989):Wage losses are significantly influenced by the length of joblessness (elasticity around -0.1). Earnings losses (JLS, 1993: high-tenure displaced workers long-term earnings losses averaging 25 % per year six years after displacement). Couch and Placzek (2010): estimates are roughly half those found for

• Pennsylvania. Greater among unemployment insurance recipients.

Farber(2011): full-time job losers who find new full-time jobs earned 11 percent less.

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 7 / 21

SLIDE 8

Introduction

Previous Literature: transferability of human capital

Most losses result from the loss of accumulated firm-specific human capital (Lefranc, 2003). Unskilled workers benefit from being attached to a particular firm while skilled workers benefit from the acquisition of transferable skills (Dustmann and Meghir, 2005). Impact of general skills and firm-specific skills to the wage growth. This allows them to find that longer lasting matches are characterized by high wage growth in the first five years and higher wages on average (Amann and Klein, 2012). The task-specific human capital explains up to 52% of overall wage growth over the career. Wage losses of displaced workers will be 10 percentage points larger for workers reemployed in a very distant

• ccupation (Gathmann and Schonberg, 2010).

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 8 / 21

SLIDE 9

Introduction

Our study

Two main objectives drive the investigation:

1 Follow JLS (1993) methodology to investigate the monthly earnings

losses, including transitions to zeros whenever the individuals are out

• f work.

2 Extend the JLS methodology by incorporating firm and job title fixed

effects in the monthly wage equation (excluding transitions to zeros), allowing us to estimate the monthly wage losses of displaced workers.

We decompose the monthly wage losses into their main sources using the methodology developed in Gelbach (2010) (omitted variables bias). Basically, we use the fixed effects to disentangle part of the wage loss.

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 9 / 21

SLIDE 10

Data

Data used: Quadros de Pessoal

Rich set of information available in the longitudinal matched employer-employee dataset for Portugal, on:

The collective agreement that regulates the employment contract applicable to each worker (300 negotiated per year, on average) Detailed occupational categories defined for each collective agreement (100 categories defined by each collective agreement, on average)

Job title: combination of collective agreement and professional category (around 30,000 per year) All the population - covers all personnel working for an establishment Very rich in worker and firm specific information (gender, age, schooling, region, industry, firm size) Period: 1997-2008

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 10 / 21

SLIDE 11

Data

Data used: Variables and definitions

Displaced: all workers who separate from a dying firm in a given year. Such workers are unlikely to have left as a result of their own poor performance and therefore it attenuates the curse self-selection. Non-displaced workers (the control group) includes all individuals that were employed at year t in a firm that did not close in year t+1 and the firm’s employment did not drop 30 percent or more and they were not subject to an individual dismissal. A firm closure is observed if the identification number of one firm appeared in period t but did not appear in t+1 and t+2. Monetary variables deflated with the Consumer Price Index (2008 prices) Hourly wage = (sum of 5 comp. of wages)/(sum of 2 types of hours)

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 11 / 21

SLIDE 12

Data

Data used: Sample

Reference year: 2002 Workers with at least three years of tenure by the time of the reference year (all years between 2002 and 2006). Full-time wage earners in the private non-farm sector Aged between 20 and 49 years Employed in a firm with at least 20 employees. Reference Period: 2002-2006

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 12 / 21

SLIDE 13

Empirical strategy

Empirical strategy

JLS(1993) wit = αi + γt + βXit +

m

• k≥−m

Dk

itδk + ǫit

JLS(1993) detrend estimator (with worker-specific time trends): wit = αi + ωit + γt + βXit +

m

• k≥−m

Dk

itδk + ǫit

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 13 / 21

SLIDE 14

Empirical strategy

Empirical framework: Linear wage equation with worker, firm, and job title fixed effects

wijft = αi + θf + λj + γt + βXift + ǫijft (1) wijft represents the monthly wage for each individual i, in job j, working for firm f in year t Xfit are observed time-varying characteristics of individual i in year t

Workers time-varying characteristics (age, age squared)

αi is a worker fixed effect θf is a firm fixed effect λj is a job title fixed effect γt are 18 year dummies ǫijft is assumed to follow the conventional assumptions

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 14 / 21

SLIDE 15

Empirical strategy

Estimation algorithm, Guimar˜ aes and Portugal (2010)

Controlling for worker, firm, and job title-specific effects requires the introduction of three high-dimensional fixed effects in the linear regression

• model. In matrix form:

Y = Xβ + D1λ + D2θ + D3γ + µ (2) where X is a matrix of time-varying explanatory variables and D1, D2, and D3 are high-dimensional matrices for the worker, firm and job fixed effects. The normal equations may be rewritten as     β λ θ γ     =     (X ′X)−1X ′(Y − D1λ − D2θ − D3γ) (D′

1D1)−1D′ 1(Y − Xβ − D2θ − D3γ)

(D′

2D2)−1D′ 2(Y − Xβ − D1λ − D3γ)

(D′

3D3)−1D′ 3(Y − Xβ − D1λ − D2θ)

    suggesting an iterative solution that alternates between estimation of β, λ, θ and γ.

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 15 / 21

SLIDE 16

Empirical strategy

Gelbach’s decomposition

1 base model with no fixed effects:

wit = γbase

t

+ βbaseXit +

m

• k≥−m

Dk

itδbase k

+ ǫbase

it

This equation has omitted variables bias.

2 Full model with the three fixed effects:

wijft = ˆ αi + ˆ θf + ˆ λj + γfull

t

+ βfullXit +

m

• k≥−m

Dk

itδfull k

+ ǫfull

ijft

3 Use ordinary least squares to estimate the vector of coefficients on

each covariate in the base model in a set of auxiliary models with each

• f the three covariates ˆ

αi, ˆ θf , and ˆ λj acting as the dependent variable

4 This algorithm results in decomposing the difference

δbase

k

− δfull

k

= ˆ τ α

k + ˆ

τ θ

k + ˆ

τ λ

k , for each time period k.

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 16 / 21

SLIDE 17

Main results

Empirical results

Figure : Monthly earnings loss of displaced workers due to firm closure

• ­‑700 ¡
• ­‑600 ¡
• ­‑500 ¡
• ­‑400 ¡
• ­‑300 ¡
• ­‑200 ¡
• ­‑100 ¡

0 ¡ 100 ¡

• ­‑6 ¡
• ­‑5 ¡
• ­‑4 ¡
• ­‑3 ¡
• ­‑2 ¡
• ­‑1 ¡

0 ¡ 1 ¡ 2 ¡ 3 ¡ 4 ¡ 5 ¡ 6 ¡ 95% ¡Confidence ¡interval ¡ Without ¡trends ¡ With ¡Trends ¡

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 17 / 21

SLIDE 18

Main results

Empirical results

Figure : Monthly wage loss of displaced workers due to firm closure

• ­‑300 ¡
• ­‑250 ¡
• ­‑200 ¡
• ­‑150 ¡
• ­‑100 ¡
• ­‑50 ¡

0 ¡ 50 ¡ 100 ¡

• ­‑6 ¡
• ­‑5 ¡
• ­‑4 ¡
• ­‑3 ¡
• ­‑2 ¡
• ­‑1 ¡

0 ¡ 1 ¡ 2 ¡ 3 ¡ 4 ¡ 5 ¡ 6 ¡ 95% ¡Confidence ¡interval ¡ Without ¡trends ¡ With ¡Trends ¡

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 18 / 21

SLIDE 19

Main results

Empirical results: The empirical distribution of wages pre-displacement (reference year D0)

,0005 ,001 ,0015 ,002 Density 1000 2000 3000 Real total monthly Displaced Non-displaced

kernel = epanechnikov, bandwidth = 50,3236

Kernel density estimate

,001 ,002 ,003 Density

• 500

500 Worker permanent heterogeneity Displaced Non-displaced

kernel = epanechnikov, bandwidth = 23,2262

Kernel density estimate

(a) Monthly wage distribution (b) Worker permanent heterogeneity

,0005 ,001 ,0015 ,002 ,0025 Density

• 500

500 Firm permanent heterogeneity Displaced Non-displaced

kernel = epanechnikov, bandwidth = 24,7446

Kernel density estimate

,001 ,002 ,003 ,004 ,005 Density

• 500

500 Job title permanent heterogeneity Displaced Non-displaced

kernel = epanechnikov, bandwidth = 10,3184

Kernel density estimate

(c) Firm permanent heterogeneity (d) Job title permanent heterogeneity

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 19 / 21

SLIDE 20

Main results

Empirical results: The empirical distribution of wages of displaced workers: pre- and post-displacement

,0005 ,001 ,0015 ,002 Density 500 1000 1500 2000 2500 3000 Real total monthly Pre-displacement Post-displacement

kernel = epanechnikov, bandwidth = 59,0026

Kernel density estimate

,0005 ,001 ,0015 ,002 ,0025 Density

• 500

500 Worker permanent heterogeneity Pre-displacement Post-displacement

kernel = epanechnikov, bandwidth = 26,0663

Kernel density estimate

(a) Monthly wage (b) Worker permanent heterogeneity

,0005 ,001 ,0015 ,002 ,0025 Density

• 500

500 Firm permanent heterogeneity Pre-displacement Post-displacement

kernel = epanechnikov, bandwidth = 27,1828

Kernel density estimate

,001 ,002 ,003 ,004 Density

• 500

500 Job title permanent heterogeneity Pre-displacement Post-displacement

kernel = epanechnikov, bandwidth = 17,5224

Kernel density estimate

(c) Firm permanent heterogeneity (d) Job title permanent heterogeneity

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 20 / 21

SLIDE 21

Main results

Empirical results: Gelbach’s decomposition

Table : Decomposition of the wage loss - displaced workers due to firm closure

Period Base Full relative OLS OLS Worker Firm Job title to displacement monthly wage monthly wage δbase

k

− δfull

k

fixed effect fixed effect fixed effect D−6 − D0

• 319.4

5.4

• 324.9
• 150.4
• 207.9

34.2 D1 − D6

• 500.6
• 6.8
• 493.7
• 166.8
• 274.8
• 52.2

∆

• 181.1
• 12.3
• 168.8
• 16.4
• 66.9
• 86.3

Results in percentage D−6 − D0

• 39.0

0.7

• 39.7
• 18.4
• 25.4

4.2 D1 − D6

• 61.1
• 0.8
• 60.3
• 20.4
• 33.6
• 6.4

∆

• 22.1
• 1.5
• 20.6
• 2.0
• 8.2
• 10.5

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 21 / 21

SLIDE 22

Robustness checks

Robustness checks

Sensitivity of losses to definition of displacement: Firm closure; Collective dismissals and Individual dismissals Sensitivity of losses to comparison group

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 22 / 21

SLIDE 23

Conclusions

Conclusions

There are severe and persistent earnings losses of workers displaced due to firm closure (51 percent of the pre-displacement wages for firm closures), six years after the separation event. The allocation into lower-paid job titles plays the most important role in explaining the wage losses of displaced workers, accounting for half

• f the total average wage loss in the case of firm closure.

Sorting into firms also plays a significant role for workers displaced through firm closures, accounting for 40 percent. Retraining programs: Severe losses in the returns to the job-title may represent a job downgrading due to depreciation

• f specific human capital

Job search assistance programs and mandatory pre-notification: Losses related with the firm fixed effect may mean that a worker is moving from a ”good” match to a ”bad” match

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 23 / 21

SLIDE 24

Appendix

Previous Literature

Earnings losses (JLS, 1993: high-tenure displaced workers long-term earnings losses averaging 25 % per year six years after displacement). Couch and Placzek (2010): estimates are roughly half those found for

• Pennsylvania. Greater among unemployment insurance recipients.

Farber(2011): full-time job losers who find new full-time jobs earned 11 percent less.

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 24 / 21

SLIDE 25

Appendix

Previous Literature: Heterogenous workers and firms - heterogeneous effects

Hijzen et al (2010): In contrast to JLS, income losses are driven mainly by non-employment spells rather than by wage losses. von Wachter et al., 2009 (in California, in the 1990s, workers with a college degree had smaller earnings losses) Old have larger losses (ChanStevens, 2001); Young have similar long-term earnings losses (Kletzer and Fairlie, 2003) Europe: large earnings losses (Bender et al ,2002 and Lefranc, 2003); reduced earnings losses (Burda Mertens (2001), Lehmann et al (2005) and Hijzenetal(2010) but... A long period of non-employment will have a large penalty in earnings (Gregory Jukes (2001), Bender et al (2002) and Abbring et al (2002).

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 25 / 21

SLIDE 26

Appendix

Previous Literature: Heterogenous workers and firms - heterogeneous effects

Hijzen et al. (2010): In contrast to JLS, income losses are driven mainly by non-employment spells rather than by wage losses. von Wachter et al., 2009 (in California, in the 1990s, workers with a college degree had smaller earnings losses) Old individuals have larger losses (ChanStevens, 2001); Young have similar long-term earnings losses (Kletzer and Fairlie, 2003) Europe: large earnings losses (Bender et al., 2002 and Lefranc, 2003); reduced earnings losses (Burda Mertens (2001), Lehmann et al. (2005) and Hijzenetal(2010) but... A long period of non-employment will have a large penalty in earnings (Gregory Jukes (2001), Bender et al (2002) and Abbring et al (2002).

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 26 / 21

SLIDE 27

Appendix

Previous Literature: transferability of human capital

Most losses result from the loss of accumulated firm-specific human capital (Lefranc, 2003). Unskilled workers benefit from being attached to a particular firm while skilled workers benefit from the acquisition of transferable skills (Dustmann and Meghir ,2005). Impact of general skills and firm-specific skills to the wage growth. This allows them to find that longer lasting matches are characterized by high wage growth in the first five years and higher wages on average (Amann and Klein, 2012). The task-specific human capital explains up to 52% of overall wage growth over the career. Wage losses of displaced workers will be 10 percentage points larger for workers reemployed in a very distant

• ccupation (Gathmann and Schonberg, 2010).

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 27 / 21

SLIDE 28

Appendix

Sample construction

Table : Displacement events in the reference period, 2002-2006

Firm Year closure 2002 2591 2003 2121 2004 2008 2005 3100 2006 1579 Total 11,399

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 28 / 21

SLIDE 29

Appendix

Sample construction

Table : Displaced workers resulting from firm closure, by period relative to displacement, 1997-2008

Firm Year closure D−6 3819 D−5 4170 D−4 4999 D−3 8180 D−2 7662 D−1 7354 D0 11399 D1 4465 D2 6665 D3 6230 D4 4252 D5 3045 D6 1759

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 29 / 21

SLIDE 30

Appendix

Sample construction

Table : Sample composition, 1997-2008

Displaced Firm Collective Individual Year Non-displaced closure dismissal dismissal 1997 222576 7379 20503 15508 1998 242560 7764 21812 17069 1999 274808 9249 25566 20056 2000 308367 9547 26000 20485 2002 308006 11312 31455 24524 2003 247774 7621 21864 18027 2004 241190 7374 20039 16722 2005 242018 7576 20373 16675 2006 235030 6903 18734 16420 2007 226502 8012 22489 17613 2008 262536 8810 24432 18794 Total 2,811,367 91,547 253,267 201,893

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 30 / 21

SLIDE 31

Appendix

Descriptives

Table : Descriptive statistics in reference year (2002)

Firm Non-displaced closure Age (years) 37 35 Tenure (years) 11 10 Female 40 46 Total monthly wage (2008 euros) 1136 819 Minimum monthly wage (2008 euros) 408 408 Hourly wage (2008 euros) 2,08 1,49 Education (percentages): Less than basic school 1 1 Basic school 23 30 Preparatory 23 33 Lower Secondary 19 15 Upper Secondary 23 15 College 11 6 Firm size (no. co-workers) 1460 195 Industry (percentages): Manufacturing 41 60 Construction 7 13 Wholesale and retail trade 20 17 Transports 10 2 Finance and business services 13 7 Education and Health 9 1

• No. Observations

308,006 11,312

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 31 / 21

SLIDE 32

Appendix

Empirical results: Gelbach’s decomposition

Table : Decomposition of the wage loss - displaced workers due to firm closure

Period Base Full relative OLS OLS Worker Firm Job title to displacement monthly wage monthly wage δbase

k

− δfull

k

fixed effect fixed effect fixed effect checksum D−6

• 270.2

17.6

• 287.9
• 135.7
• 212.9

62.3

• 1.6

D−5

• 278.2

8.8

• 287.0
• 130.4
• 213.4

58.0

• 1.2

D−4

• 295.4
• 2.5
• 292.8
• 152.8
• 181.0

41.7

• 0.7

D−3

• 298.7
• 18.4
• 280.3
• 141.9
• 184.0

46.2

• 0.7

D−2

• 322.0

11.1

• 333.1
• 151.0
• 214.8

33.2

• 0.5

D−1

• 395.5
• 10.5
• 384.9
• 168.1
• 220.5

4.0

• 0.3

D0

• 376.1

32.1

• 408.2
• 172.8
• 229.1
• 6.1
• 0.1

D1

• 421.2
• 12.7
• 408.5
• 137.4
• 237.8
• 33.2
• 0.1

D2

• 492.6
• 6.1
• 486.5
• 178.8
• 253.0
• 54.6

0.0 D3

• 514.6

3.9

• 518.4
• 185.8
• 264.8
• 67.9

0.0 D4

• 574.7
• 10.1
• 564.6
• 198.8
• 300.5
• 65.4

0.1 D5

• 508.0

19.3

• 527.3
• 180.5
• 290.6
• 56.5

0.3 D6

• 492.3
• 35.4
• 456.9
• 119.3
• 302.3
• 35.4

0.1 D−6 − D0

• 319.4

5.4

• 324.9
• 150.4
• 207.9

34.2

• 0.7

D1 − D6

• 500.6
• 6.8
• 493.7
• 166.8
• 274.8
• 52.2

0.1 ∆

• 181.1
• 12.3
• 168.8
• 16.4
• 66.9
• 86.3

0.8 Results in percentage D−6 − D0

• 39.0

0.7

• 39.7
• 18.4
• 25.4

4.2

• 0.1

D1 − D6

• 61.1
• 0.8
• 60.3
• 20.4
• 33.6
• 6.4

0.0 ∆

• 22.1
• 1.5
• 20.6
• 2.0
• 8.2
• 10.5

0.1

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 32 / 21

SLIDE 33

Appendix

Decomposition of wage variability

Contribution of the ten components to the real hourly wages: Worker fixed effects: 36.0%

Unobserved component: 21.0% Observed component (gender and education): 15.0%

Firm fixed effects: 28.7%

Unobserved component: 14.6% Observed component (region, capital ownership, and industry): 14.0%

Job title fixed effects: 9.7%

Unobserved component: 1.9% Observed component (occupation and collective agreement): 7.9%

Individual time-varying characteristics: 17.4%

Time: 6.2% Time-varying observable characteristics of workers (age and seniority): 2.9% Time-varying observable characteristics of firms (size): 5.3%

Raposo, Portugal, Carneiro (2013) Decomposition of wage losses Paris, OECD, May 2013 33 / 21